Real Exchange Rate Volatility on the Short- and Long-Run

Trade Dynamics in Colombia

José J. Cao-Alvira∗†

Abstract

The short- and long-run implications of real exchange rate volatility on Colombian bilateral

trade commodities and non-commodities with its major trade partners are analyzed from the per-

spectives of the Marshall-Lerner condition, a cointegration relation with other aggregate variables

and the J-curve hypothesis. Long-run equilibrium on the Colombian bilateral balance of trade

with a country is more common when the trade volume is denominated in terms of one of the

world's main currencies; as is the case of commodity trade and trade with a country which its

national currency is one of these currencies. No evidence of the J-curve was found.

JEL Classi�cation: C22, F14, F31, N76

Keywords: bilateral balance of trade; real exchange rate �uctuations; Colombia

1 Introduction

Recent experiences of prolonged and successive appreciations of the Colombian peso have raised

credible concerns on the stability of the emerging country's competitiveness and the current surplus

in the balance of trade. Concerns on the permanence of the country's competitiveness strengths vis-

à-vis improvements in its terms of trade are readily apparent since any international competitiveness

gain in a Colombian industrial sector, due to enhancements in unit labor requirements and/or wage

di�erentials, can be o�set by an overvalued national currency. The validation of the concerns regarding

an appreciated Colombian peso and its e�ect on the stability of the current trade surplus are a greater

challenge to assess. This is due to the heterogeneous categories of goods in the Colombian balance of

trade and their individual and possibly distinct short- and long-term reactions to changes in the terms

of trade.

The 2003-2011 nine years period was one of vertiginous appreciation of the Colombian currency; e.g.

a 40.1% total real appreciation or a 6.3% yearly rate. This appreciation of the peso was accompanied

∗Associate Professor of the Finance Department and Chair of the Graduate School of Business Administration,University of Puerto Rico - Río Piedras†I gratefully acknowledge the excellent research assistance of Gabriela Galindez and Melissa Gonzalez from the UPR-

RP, and the data support of Jorge E. Cabrales from the Colombian SIEX-DIAN and the EIAM-USA. The author received�nancial support from the PII-2013 Research Grant of the Business School at the UPR-RP.

1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011−2.5

0

2.5x 10

10

1998 2000 2002 2004 2006 2008 201050

100

150

NX−Commodities NX−NonCommodities NX−All Goods Real Exchange Rate Index

Figure 1: Colombian real exchange rate index (right axis), and net exports of all goods, commoditiesand non-commodities (left axis) with its main trade partners for the 1998-2011 period. Source: CentralBank of Colombia.

by a positive trade balance. Figure (1) illustrates the volume of Colombian total net exports with its

main trade partners1 and the corresponding value of its real exchange rate index for the 1998-2011

period.2 The diagram in Figure (1) also contains the series of net exports decomposed in commodity

and non-commodity goods. Figure (1) shows that since the year 1999 Colombia has been a net exporter

of commodities and a net importer of non-commodities, and comparing the magnitudes of the net-

export volumes of both types of goods, it is evident that the surplus in total trade results from the

superior increases in the volume of the trade surplus of commodities relative to the trade de�cit of

non-commodities. The increases in net exports of Colombian commodities during this period is with

certainty a response to the increases in world prices of commodities, which for this period increased as

1Colombia's main international trade partners are Germany, Belgium, Brazil, China, Ecuador, U.S.A., Netherlands,Italy, Japan, Mexico, Panama, Peru, U.K., Dominican Republic, Switzerland and Venezuela. Sources and descriptionsof the data are provided in Section 3 of the paper.

2The Real Exchange Rate Index published by the Banco de la República of Colombia is used (1994=100). An increasein the Index indicates a real devaluation of the Colombian peso. A detailed explanation of the methodology used toconstruct the Index can be found in Huertas (2003).

2

an aggregate at a yearly rate of 11.3% (and of 17.3% if only oil is to be considered). 3

A visual assessment on the behavior of the time series in Figure (1) suggests the existence of a

degree of correlation between the exchange rate and net export dynamics, that is the most apparent for

the trade of non-commodities. Not readily evident from Figure (1) is that the volume of the Colombian

external sector with its main trading partners is dominated by the import of non-commodities, with an

annual average value of $13,632 million U.S. dollars. Colombian annual exports of non-commodities to

these commercial partners average $9,346 million USD, yielding the trade de�cit of non-commodities

observed in Figure (1). Annual Colombian commodity exports average to $7,498 million USD, with

the vastest of these corresponding to oil exports to the United States. Colombian commodity imports

average to an annual $400 million USD. As previously observed in Figure (1), the end result of adding

the Colombian commodity trade surplus and the trade de�cit non-commoditis is a positive balance of

trade.

These characterizing traits of the country's external sector oblige an analysis on the Colombian

balance of trade to incorporate a level of disaggregation in the trade volume that allows for individual

analyses on commodity and non-commodity commerce. The composition of the volume of goods in

the balance of trade of Colombia ultimately a�ects the impact that exchange rate volatility has on

the country's trade balance, that is because non-commodity trade is likely to be denominated in the

currency of the exporting country, while commodity trade is often priced in terms of the world's main

currencies, mainly the U.S. dollar; see Boughton & Branson (1988) and Roberts & Schlenker (2010).

The currency in which an international trade transaction is denominated, argues Wilson (2001), can

limit the capacity to transfer prices among agents resulting from exchange rate variations.

The main question raised in this paper is whether an improvement in the Colombian terms of trade

will ultimately detriment the country's bilateral trade balances of commodity and non-commodity

goods. Addressing this question, a dynamic adjustment mechanism is used to analyze the short-run

and long-run responses of Colombian international trade to real exchange rate volatility. The research

is performed from the perspectives of the Marshall-Lerner condition, a cointegration analysis among

3Note: The data sources from which the growth rates were calculated are the All Commodity Price Index and CrudeOil Price Index, available at the data appendix of the IMF Commodity Market Review.

3

trade variables and the J-curve hypothesis.

The next section of this paper reviews the existing literature on the e�ects of real exchange volatility

on short- and long-run trade balance dynamics. Section 3 provides a characterization of the bilateral

trade of commodities and non-commodities of Colombia with its main trading partners and details

the sources of the data used for the study. Section 4 presents the model used for analysis. Section

5 presents the empirical �ndings and Section 6 concludes summarizing the results of study and some

policy considerations.

2 Literature Review

The scienti�c literature o�ers no conclusive answer on whether exchange rate manipulations are

e�cient public policy instruments for the correction of current account de�cits. Signi�cant works

argue on the possible existence of distinct response behaviors for the balance of trade to exchange

rate variations over the short-run and the long-run horizons. Trade data analysis over a long-run

horizon is generally focused on the Marshall-Lerner condition (or M-L condition) and the search for a

cointegration relation on the time series of balances of trade and exchange rates.

The M-L condition states, see Goldstein & Khan (1985) and Argy (1994), that exchange rate

volatility yields changes in a country´s trade �ows ultimately a�ecting its current account balance,

and the price elasticities of imports and exports are what determine the magnitude of the changes in

the current account. The M-L condition is said to hold when the sum of the absolute values of these

elasticities is greater than unity (Bahmani-Oskooee, 1998).

Cointegrated time series are those that move together in a path towards long-run equilibrium.

Researchers, see for instance: Bahmani-Oskooee & Goswami (2003), Halicioglu (2008), Bahmani-

Oskooee & Hajile (2009) and Shahbaz et al. (2012), test for cointegration among trade balance and

exchange rate time series assuming a dynamic adjustment mechanism on the trade balance. The

models of choice for these enviroments are Pesaran et al. (2001) Autoregressive Distributed Lag

(ARDL) cointegration model and adaptations of the Engle & Granger (1987) Error Correction (EC)

model. These cointegration models tests on whether the short-term dynamics in the systems are in

e�ect in�uenced by the deviation from equilibrium and if there is a tendency to correct any short-term

4

error.

The J-curve hypothesis has recently become the work-horse for academics and practitioners alike

for analyzing the short-run dynamics of the balance trade and exchange rate relation. The J-curve

hypothesis, initially introduced in Junz & Rhomberg (1973) and Magee (1973), references the theorized

behavior that follows a country's balance of trade following a real devaluation, initially registering a

de�cit followed by a surplus, resembling a �J�. The value of trade contracts negotiated prior and after

the devaluation is to be a�ected by the new terms of trade. The assumed initial de�cit in the balance

of trade after a devaluation is responding to the higher prices of imports faced by the home country.

This immediate response in the balance of trade is known the �price e�ect�, see Gupta-Kapoor &

Ramakrishnan (1999). Once trade contracts are negotiated under the new terms of trade, the volume

of home exports are expected to increase (e.g. the �volume e�ect�).

Bahmani-Oskooee & Hegerty (2010) provides an excellent and very complete review on the �ndings

of the existing empirical scienti�c literature on the J-curve. The authors' survey suggests that studies

using aggregate trade data may result in ambiguous or con�icting results, because aggregate data

conceals signi�cant movements of variables within its subsets. Bahmani-Oskooee & Hegerty (2010)

recognizes that studies employing bilateral and disaggregated data at industry or sector-speci�c level

have a higher capacity to identify the presence of a J-curve. Of the surveyed works included in Bahmani-

Oskooee & Hegerty (2010), only two addressed the presence of the J-curve in the balance of trade in

Latin American countries. These are Gomes & Paz (2005) and Bahmani-Oskooee & Hegerty (2011)

which, respectively, searched for evidences of J-curves in the trade balances of Brazil and Mexico.

Gomes & Paz (2005) found evidence of the J-curve on Brazilian trade balance using aggregate data.

Bahmani-Oskooee & Hegerty (2011) analyzed industry level bilateral trade data between the U.S. and

Mexico, and found no J-curve. The authors argue that because of the level of economic integration and

the prevalence of inter-industry trade between these countries trade �ows are found to be relatively

insensitive to the �uctuations of the Mexican peso. These studies are of particular interest to this

paper because all use comparable cointegration methods on a dynamic adjustment mechanism to test

for the existence of the J-curve, and when three important similarities between Mexico, Brazil and

5

Colombia are considered. First, the three countries are members of the LAC-7 group, meaning that

they belong to the seven largest Latin American economies.4 Second, the three countries experienced

in the previous decade a deep and persistent degree of openness in their economies.5 Third, the three

countries have an external sector where commodity exports are of great importance. International

commerce of commodities can have a bu�er e�ect on the balance of trade to exchange rate volatility.

As previously discussed, the degree to which prices are transferred and net exports a�ected from

exchange rate variations during the international commercial trade of goods depends on the price

elasticity of the imports and exports. Commodities are characteristically of low price elasticity, which

decreases the possibility of price transfers; see Dwyer, Kent & Pease (1994). Section 3 discusses in

detail the external sector of Colombia and the bilateral trade data used in the analysis of this study.

3 Data

The countries selected for the study are Colombia's primary bilateral trade partners, based on the

criteria that these consistently ranked among the ten countries that Colombia had the highest yearly

trade volume (exports plus imports) for the 1998-2009 time period. These are Germany, Belgium,

Brazil, China, Ecuador, the U.S., the Netherlands, Italy, Japan, Mexico, Panama, Peru, the U.K.,

the Dominican Republic, Switzerland and Venezuela. Total Colombian trade with these countries

during the considered time period represents, on average, 80.1% of its total exports and 72.7% of its

total imports. These sixteen Colombian trade partners include four LAC-7 member countries and six

countries which their national currency is one of the world's main currencies, e.g. the U.S. dollar, the

Euro, the British pound and the Japanese Yen.

The data of interest for the study are the time series (i) of Colombian bilateral trade volume, disag-

gregated according to their respective Harmonized System (HS) Code Classi�cations, (ii) of Colombian

bilateral real exchange rate with each of the considered trade partners, and (iii) of the national income

of Colombia and each trade partner. All data is ultimately expressed in quarterly frequency.

4LAC-7 members are Argentina, Brazil, Chile, Colombia, Mexico, Peru and Venezuela.5According to the Openness Index, the yearly openness growth rate of Colombia for the 1998-2009 time period was

2.1%, for Brazil 3.3% and for Mexico 3.8%. Source: Heston, et. al. (2012), Penn World Table Version 7.1.

6

Table 1: Volume of Colombian Exports and Imports (yearly million USD, nominal 1998-2009)

Comm NonComFlowers Co�ee Corn Sugar Coal Oil Gold FNickel

GERX 214.03 7.78 198.24 - 0.37 - 0.80 0.01 6.83 212.80

M 0.42 5E-4 - - 0.01 - 0.40 - 0.01 786.96

BELX 101.49 0.05 71.51 - 0.91 1.54 1.10 - 26.38 213.56

M 0.43 - - - 0.08 - 0.35 - - 94.90

BRAX 74.22 1.45 1E-3 - - 14.13 53.26 - 5.38 183.19

M 22.61 - 0.62 1.25 10.92 - 9.82 - - 1,106.94

CHINX 169.74 0.01 0.81 - 0.21 3E-3 41.23 - 127.48 95.86

M 0.02 2E-4 2E-3 - 0.01 - 0.01 - - 1,448.12

ECUX 33.58 0.06 0.19 0.10 10.37 0.21 22.65 - 1E-4 905.02

M - - - - - - - - - 470.54

U.S.X 5,688.5 631.49 463.6 7E-4 23.46 14.13 4,277.4 238.4 40.06 2,161.26

M 248.92 0.01 0.05 0.71 0.20 0.02 247.93 - 6E-4 5,428.98

ITAX 144.61 0.69 40.19 - 0.23 1.09 - 1.23 101.18 189.95

M 1.04 - 0.06 - 0.00 - 0.98 1E-4 - 359.41

JAPX 223.34 10.49 181.3 - 3E-3 0.04 - - 31.51 53.98

M 1.05 - - - - 0.02 1.03 - - 728.02

MEXX 55.45 0.18 1.31 0.20 5.22 15.27 33.27 - - 348.88

M 5.71 - - 0.68 0.40 - 4.63 6E-4 - 1,365.74

NETX 74.01 11.80 32.07 - 0.08 1.78 0.44 3E-3 27.84 374.63

M 2.03 0.01 1E-4 - - - 2.02 - - 156.55

PANX 34.62 0.98 0.06 0.07 0.72 0.10 32.22 0.47 - 190.30

M 1.19 - - - - - 1.19 - - 67.14

PERX 111.77 4E-3 0.08 0.32 29.00 21.27 61.09 - 0.01 430.94

M 11.04 0.01 6.71 0.07 4.25 - - 307.92

U.K.X 101.43 33.57 39.64 - 0.03 2.90 23.29 2.00 - 244.22

M 8.68 1E-4 0.01 - - - 8.67 - - 204.46

DOMX 209.37 4E-3 - 5E-3 2.29 0.07 207.0 - - 151.12

M 0.00 - - - - 3E-3 - - 3.73

SWIX 206.67 0.95 - - 0.02 1.36 77.30 123.7 - 136.80

M 0.20 - - - - - 0.20 - - 220.22

VENX 55.47 4.15 0.64 0.81 38.77 4.85 6.25 - - 2,331.7

M 97.05 - 6E-4 0.12 0.01 1E-3 96.42 0.50 - 882.83

The source of Colombian bilateral trade data is the Statistical System of International Commerce at

the Bureau of National Taxes and Tari�s of Colombia (SIEX at the DIAN, for their respective Spanish

abbreviations), and is expressed in �Free on Board� (or FOB) and in monthly nominal U.S. dollars.

The monthly data on the volume of bilateral trade is converted to quarterly frequency in order to make

it comparable with the other data utilized in the analysis. The year 1998 is chosen as the start of the

analyzed period because it is from this date that the SIEX at the DIAN stores and makes publicly

available disaggregated bilateral trade data for Colombia. The analyzed time period extends until the

end of calendar year 2009 because then Colombia and Venezuela abruptly halted their international

7

trade due to diplomatic con�icts, ref: The Economist (September 10, 2009). Diplomatic relations were

restored on August 2010, ref: Bloomberg (August 11, 2010).

The disaggregated bilateral trade data is classi�ed as commodities or non-commodities. Commodity

goods are identi�ed as those traded goods which their prices are determined in international commodity

exchange markets or are �xed seasonally upon supply and demand dynamics. These goods are �owers

and �ower bulbs, toasted and untoasted ca�einated and deca�einated co�ee, corn, unre�ned cane

sugar, coal, crude oil and bituminous minerals, raw gold and ferronickel.6 Non-commodity goods are

identi�ed as those that remain in the bilateral trade time series after commodity goods are identi�ed

and removed from the series. Table (1) contains the annual average values, in millions of nominal USD,

of Colombian bilateral exports and imports of commodities and non-commodities with its main trade

partners for the 1998-2009 time period. Data on the volume of the bilateral trade of commodities is

further dissaggregated in Table (1) using the previously described categories of traded goods.7

Colombia's primary commercial partner is the U.S.; this country is the destination of 41.1% of its

exports and the source of 28.7% of its imports. Colombia's second most important trade partner is

Venezuela, with whom shares a 2,219 km frontier, receiving 11.1% of Colombian exports and supplying

4.88% of imports. China is the third principal Colombian trade partner, despite being the destination

of only one percent of Colombian exports and this country supplies 8.1% of imports. The main

categories of exported goods, in descending order by volume, are crude oil, vegetable products, and

mineral products to the U.S., and textile products to Venezuela and the U.S.. Main imports, also

in descending order, are machinery and electrical equipment, chemical products, and transportation

equipment from the U.S., machinery and electrical equipment from China, and machinery and electrical

equipment from Mexico.

The bilateral real exchange rate between Colombia and country j at time period t (or RERjt ) is

calculated according to RERjt = (P colt ·NERjt )/P

jt ; where P

colt and P jt are the consumer price indexes

of Colombia and country j at time t, and NERjt is the nominal exchange rate between Colombia and

country j at time t. National income of country j at time t, Y jt , is measured using the quarterly real

6Appendix A provides a detailed description of those goods considered in the study as commodities with theirrespective HS Code Classi�cations.

7Data on the bilateral trade of noncommodities, by major types of goods and services, is presented in Appendix B.

8

gross domestic product. Bloomberg is the source of quarterly data on bilateral exchange rate, and

national price indexes and income statistics.

4 Modeling Environment

Equations (1), (2) and (3) contain the modeling environment for Colombian commodity and non-

commodity bilateral trade with partner j. The setting folows the dynamic adjustment mechanisms

in the Autoregressive Distributed Lag (ARDL) cointegration model of Pesaran et al. (2001) and the

Engle & Granger (1987) Error Correction (EC) model, as advanced by Halicioglu (2008) and Bahmani-

Oskooee & Hajile (2009) for bilateral trade scenarios. Equation (1) is the reduced form of the bilateral

trade balance model. tbji,t is the natural logarithm of Colombian export to import ratio of commodities

and non-commodities, i.e. i = {com, noncom}, with country j at time t. rerjt is the natural log of

the real exchange rate between the Colombian and country j's currencyat at time t. An increase in

rerjt represents a real devaluation of the Colombian peso and, with it, a detriment of Colombia's terms

of trade. yjt and ycolt are the natural logs of country j's and Colombia's national income at time t,

respectively, measured using each country's real gross domestic product. The model is set in quarterly

frequency.

tbji,t = αji + λji · rerjt + βji · y

jt + βcol,ji · ycolt + εji,t (1)

λji is the real exchange rate elasticity of the Colombian trade balance of types of goods i with country

j and is considered being the long-run response of tbji,t to bilateral real currency devaluations. If a

real devaluation of the Colombian currency is believed to increase the country's exports and decrease

imports, then λji is expected to be positive. The Marshall-Lerner condition holds when the elasticity

coe�cient λji is greater than one with statistical signi�cance. βcol,ji and βji respectively denote the

average percentage change in tbji,t as a response to a percentage increase in Colombian and country j's

real income. Either, a positive or negative sign for βcol,ji and βji is sustained by the literature. βcol,ji

could be positive if an increase in Colombian real income leads also to an increase in imports of type

i goods. Yet, if the increase in real income is due to increases in national production of good i import

substitutes then βcol,ji is expected to have a negative sign; see Halicioglu (2008). Similar arguments

9

apply for the expected sign of βji .

Equation (2) contains the ARDL representation of Equation (1). Equation (2) considers a dynamic

adjustment mechanism on the Colombian bilateral trade balance, where the di�erences in tbji,t are

regressed against the lagged values of tbji,t, rerjt , y

jt and ycolt , and their n0, n1, n2 and n3 order

lagged di�erences. Optimal lag lengths n0, n1, n2 and n3 are chosen based on the Akaike Information

Criterion. The hypothesis of a J-curve pattern on the Colombian bilateral trade balance with country

j is supported when λji,k assumes negative values at lower lags and positive values at higher. A

cointegration relation between tbji,t, rerjt , y

jt and y

colt , i.e. moving in a path together towards long-run

equilibrium, can be tested under the null hypothesis of no cointegration, e.g. Ho : δj0 = δj1 = δj2 =

δj3 = 0. Considering that the F-test for testing the null has a non-standard distribution, Pesaran et.

al. (2001) computes lower and upper bounds of critical values, assuming all variables are I(0) or I(1).

If the F-statistic is higher [lower] than the upper [lower] bound, then reject [do not reject] the null. If

the F-statistic is between the bounds, then the test is inconclusive.

∆tbji,t = θji +

n0∑k=1

ϕji,k∆tbji,t−k +

n1∑k=0

λji,k∆rerjt−k +

n2∑k=0

βji,k∆yjt−k +

n3∑k=0

βcol,ji,k ∆ycolt−k

...+ δji,0 · tbji,t−1 + δji,1 · rer

jt−1 + δji,2 · y

jt−1 + δji,3 · y

colt−1 + uji,t (2)

On Equation (3), the ARDL model in Equation (2) is reformulated into a general Error Correction

model. ECjt−1 is the error-correction term and is constructed using the lagged residual of the reduced

form model in Equation (1); i.e. ECji,t−1 = tbji,t−1 −[αji + λji · rer

jt−1 + βji · y

jt−1 + βcol,ji · ycolt−1

]. A

statistically signi�cant negative value of coe�cient φji can also be used to test the null hypothesis of

no cointegration among the variables. Kremers et. al. (1992), Bahmani-Oskooee & Goswami (2003)

and Halicioglu (2008) are examples in the literature where the statistical signi�cance of φji is used

to directly test for cointegration or to conclusively test variable cointegration for those cases where

the calculated F-statistic of the ARDL model yield an inconclusive result. The Akaike Information

Criterion is used to choose the optimal lag lengths n0, n1, n2 and n3 for Equation (3).

∆tbji,t = θji +

n0∑k=1

ϕji,k∆tbji,t−k +

n1∑k=0

λji,k∆rerjt−k +

n2∑k=0

βji,k∆yjt−k +

n3∑k=0

βcol,ji,k ∆ycolt−k

...+ φji · ECji,t−1 + uji,t (3)

10

The model estimation is evaluated based on its overall �t, stability of coe�cients, sequential cor-

relation of the residuals and misspeci�cation. Overall �t is measured using the adjusted R-squared of

the estimation. The CUSUM and CUSUMSQ tests, originally developed in Brown et. al. (1975) and

based on the recursive regression residuals using their cumulative sum and their cumulative sum of

squares, are used to assess the estimated coe�cients stability. Stability is not implied by cointegration,

as stated in Bahmani-Oskooee and Brooks (1999). If the graphical representations of these statistics

are within the critical bounds of 5% signi�cance, then the coe�cients of the regression are stable.

Serial correlation is tested under the null of no serial correlation using a Lagrangian Multiplier statis-

tic, which has χ2 distribution with one degree of freedom. The model misspeci�cation is tested using

Ramsey's RESET test under the null of a not misspeci�ed model, and also is distributed according to

χ2 with one degree of freedom.

5 Empirical Findings

Tables (2) and (3) contain the coe�cient estimates of the real exchange rate regressors λji for the

long-run reduced form model in Equation (1), and for the lagged di�erences of the real exchange rate

regressors{λji,k

}4

k=0and the lagged error-correction term ECji,t−1 of the short-run dynamic adjustment

error-correction model in Equation (3). Tables (2) and (3) also contain the diagnostic statistics that

test the appropriateness of the model described in Equations (1)-(3) for analyzing Colombian bilateral

trade data. These are the CUSUM and CUSUMSQ test statistics on the stability of the residuals of

the optimal models, the χ2 statistic from the Lagrangian Multiplier on a test for serial correlation

on the residuals of the model estimation, the χ2 statistic from the Ramsey's RESET test for model

misspeci�cation, and the adjusted R square indicating the overall �t of the estimation.

The J-curve hypothesis, tested by inspecting the presence of statistically signi�cant negative coef-

�cients for{λji,k

}4

k=0at early lags and positive at later lags, is not sustained for any of the analyzed

bilateral trade scenarios. Although no instances were found con�rming the J-curve hypothesis, indi-

vidual scenarios arouse where either Colombian bilateral net exports initially decrease following a real

devaluation of the Colombian currency (e.g. the price e�ect of a devaluation), or where Colombian

net exports increased in the longer run trailing a devaluation (e.g. the volume e�ect). As expected,

11

Table 2: Exchange Rate Coe�cient for Colombian Commodity Trade

i =comm

λji,0 λj

i,1 λji,2 λj

i,3 λji,4 λj

i ECji,t−1

LMreset

csumcsum2

R2adj

GER-1.95 -0.68 -0.74 0.1 S

0.57(1.3) (1.6) (3.7)*** 0.2 S

BEL-0.68 -3.5 -0.94 1.1 S

0.40(0.2) (3.2)*** (4.5)*** 0.9 S

BRA-0.66 5.46 -0.57 2.5 S

0.50(0.2) (2.2)** (2.9)*** 0.3 S

CHIN-6.4 5.03 -0.68 1.8 U

0.48(0.2) (0.6) (3.2)*** 0.1 U

ECU-0.52 2.22 -0.76 1.9 S

0.36(0.3) (2.9)*** (3.4)*** 0.0 S

U.S.0.72 -2.03 0.12 -2.11 2.2 S

0.70(0.5) (1.6) (0.2) (5.9)*** 0.0 S

ITA1.29 5.4 1.51 -3.74 2.36 0.56 -1.86 0.4 S

0.51(0.2) (0.8) (0.2) (0.6) (0.4) (0.4) (3.0)*** 0.0 S

JAP1.94 2.43 0.42 -3.14 4.41 -0.37 -1.15 3.4* S

0.54(0.5) (0.6) (0.1) (0.8) (1.2) (0.5) (2.5)** 2.5 S

MEX-2.95 11.98 -15.2 -12.64 -4.76 -2.74 -0.57 1.3 S

0.49(0.4) (1.8)* (2.0)* (1.7)* (0.7) (1.1) (2.5)** 1.1 S

NET-18.07 -18.8 -0.67 0.7 U

0.50(1.4) (4.4)*** (2.0)** 0.0 U

PAN50.28 20.96 0.45 13.34 -47.95 22.27 -1.49 1.3 U

0.51(1.0) (0.4) (0.0) (0.2) (1.0) (2.1)** (2.7)*** 5.2** S

PER-5.28 -2.33 -8.45 6.6 1.46 -1.55 -0.68 1.2 S

0.35(1.1) (0.5) (1.7) (1.0) (0.2) (1.4) (1.8)* 0.2 S

U.K.7.61 4.96 -8.5 5.09 -2.94 3.68 -0.73 7.2*** S

0.55(1.7)* (1.0) (2.0)* (1.0) (0.6) (1.6) (2.9)*** 1.0 U

DOM0.78 -3.67 14.13 -5.3 -17.97 -1.98 -1.31 0.2 U

0.41(0.1) (0.3) (1.1) (0.4) (1.3) (0.6) (1.8)* 1.2 U

SWI9.23 -1.05 -0.76 2.7* S

0.47(1.5) (1.0) (2.2)** 0.0 U

VEN0.36 1.55 -7.02 3.43 1.36 1.93 -1.06 12.4*** S

0.27(0.1) (0.4) (1.7) (0.9) (0.4) (1.7)* (2.2)** 0.7 S

Colombian non-commodity trade proves being more short-run responsive to exchange rate volatility

than commodity trade. Following a home-currency devaluation, signi�cant short term declines are

present for Colombian bilateral non-commodity trades with Germany, Belgium, China, the U.K. and

Switzerland. Combined, these �ve partners are the destination of 14% of Colombian non-commodity

exports and the source of 20% of its imports. This short-run e�ect never occurred for the cases analyzed

of bilateral trade of commodities.

Colombian bilateral trade of commodities with Brazil, Ecuador, Panama and Venezuela showed

signi�cant long-run improvements after the real devaluation of the Colombian currency. Three per-

12

cent (3%) of Colombian commodity exports are destined to these countries and 30% of its commodity

imports are sourced from them. The Marshall-Lerner condition holds for the cases of Colombian

bilateral commodity trade with Brazil and Panama, that is one percent (1%) of total commodity

exports and six percent (6%) of imports. A devaluation of the Colombian peso is found to be associ-

ated with long-run increases in the country's bilateral trade balance of non-commodities with Brazil,

Ecuador, the U.S., Italy, Japan, Mexico, the Netherlands, Panama, Peru and Venezuela. These coun-

tries are the destination of 77% of Colombian non-commodity exports and the source of 80% of imports.

The Marshall-Lerner condition holds for the non-commodity trade with Brazil, the Netherlands and

Venezuela. These countries correspond to 31% of total Colombian non-commodity exports and 16%

of imports.

Trade scenarios are observed where a real devaluation of the Colombian peso causes short-run im-

provements in the country's trade balance or long-run detriments. A devaluation of the Colombian

currency, on average, causes short-run increases in the bilateral commodity trades with Mexico and

the U.K.. Commodity trade with these countries during the 1998-2009 period averages to merely 2%

of Colombian exports and 4% imports. Nevertheless, these instantaneous increases are immediately

followed by even greater decreases in the trade balance, resulting in a negative net outcome for the

bilateral trade balance in a longer horizon short-run. The long-run impact of a devaluation of the

Colombian peso in the bilateral commodity trades with Belgium and the Netherlands is negative.

Colombian commodity trade with these countries represents 2% of exports and 1% of exports. Colom-

bian bilateral non-commodity trades with Brazil, the U.S., the Netherlands, Panama and Venezuela

respond positively in the short-run to a real devaluation of the Colombian peso. These countries rep-

resent 56% of Colombian non-commodity exports and of its imports. Bilateral non-commodity trade

with Germany (2% of Colombian non-commodity exports and 6% imports) responds negatively in the

long-run to a real devaluation of the Colombian currency.

Cointegration, de�ned as the occurrence of a long-run relationship among the variables in the model,

is considered to exist provided the presence of a statistically signi�cant negative value for the error

correction coe�cient in Equation (3). The values reported in Table (2) indicate that there is presence

13

Table 3: Exchange Rate Coe�cient for Colombian Non-Commodity Trade

i = nocomm

λji,0 λj

i,1 λji,2 λj

i,3 λji,4 λj

i ECji,t−1

LMreset

csumcsum2

R2adj

GER-0.16 -0.57 -1.19 0.98 -0.74 -0.81 1.3 S

0.45(0.2) (0.9) (1.7)* (1.5) (3.7)*** (3.9)*** 0.0 S

BEL-0.42 0.5 0.29 -0.49 -1.79 -0.1 -0.88 5.0** S

0.51(0.6) (0.7) (0.5) (0.8) (2.8)*** (0.5) (4.0)*** 1.1 S

BRA2.98 0.38 0.22 0.14 1.82 2.26 0.08 0.4 S

0.32(3.8)*** (0.6) (0.4) (0.2) (2.8)*** (4.6)*** (0.3) 0.7 S

CHIN-1.82 0.29 -0.63 6.5** S

0.50(3.0)*** (0.9) (4.8)*** 0.1 S

ECU0.31 0.47 0.35 -0.76 2.7 S

0.39(0.9) (1.5) (2.2)** (4.2)*** 0.2 S

U.S.1.00 0.41 -0.05 0.88 -0.84 0.44 -0.35 7.5*** S

0.60(2.4)** (0.9) (0.1) (2.2)** (2.2)** (2.4)** (2.0)* 0 S

ITA1.07 1.04 -1.43 0.8 -0.35 0.77 -1.06 1.2 S

0.52(0.9) (1.0) (1.4) (0.8) (0.3) (3.0)*** (2.0)* 5.5** S

JAP-0.42 -0.5 0.15 -0.15 0.21 -0.8 2.1 S

0.34(0.7) (0.8) (0.2) (0.2) (1.8)* (2.6)** 1.6 S

MEX0.06 0.32 0.7 0.02 0.4 0.99 -0.28 6.8*** S

0.08(0.1) (0.5) (0.9) (0.0) (0.6) (3.7)*** (1.3) 1.1 S

NET-0.13 3.35 -1.81 1.53 -1.81 4.7** S

0.70(0.2) (3.9)*** (1.9)* (5.2)*** (6.1)*** 0.2 S

PAN1.43 -2.43 3.93 -5.66 2.69 0.89 -0.56 3.4* S

0.50(0.7)* (1.1) (1.6) (2.3)** (1.3) (1.8)* (1.6) 0.6 S

PER0.01 -0.12 0.07 -0.06 -0.22 0.32 -0.52 2.4 S

0.08(0.0) (0.2) (0.1) (0.1) (0.4) (3.3)*** (1.8)* 0.8 S

U.K.-2.34 -1.74 0.23 0.86 -0.12 -0.64 -0.83 0.5 S

0.48(1.8)* (1.5) (0.2) (0.7) (0.1) (1.2) (2.5)** 1.8 S

DOM0.16 -0.45 1.01 -0.4 -0.35 -0.42 -1.06 2.5 S

0.40(0.2) (0.5) (1.2) (0.4) (0.4) (1.2) (2.4)** 0.3 S

SWI-0.29 -2.58 -0.44 -0.36 0.2 S

0.11(0.2) (1.7)* (0.9) (2.2) 4.1** S

VEN1.21 1.36 -1.75 -0.51 -0.19 2.11 -0.71 2.9* S

0.48(2.0)** (2.1)* (2.0)** (0.6) (0.3) (6.4)*** (2.8)*** 4.8** U

of cointegration in Colombian commodity bilateral trade with all analyzed trading partners. Cointe-

gration is less common in Colombian non-commodity trade, as is observed in Table (3) . Cointegration

is not present on Colombian non-commodity trade with Brazil, Mexico, Panama and Switzerland.

The results of the diagnostic tests on the model estimation results, e.g. overall �t, stability of

coe�cients, sequential correlation of the residuals and model misspeci�cation, suggest that the model

described in Equations (1)-(3) is appropriate for analyzing Colombian bilateral trade data. The overall

�t of each of the estimated models is measured using the adjusted R-square statistic. With the excep-

14

tion of the regression equations modeling Colombian non-commodity trade with the Netherlands, Peru

and Switzerland, all estimated equations are able to explain over a third of the observed variability

(adjusted for degrees of freedom) in ∆tbji,t. The combined result of the CUSUM and CUSUMSQ tests

(both considering bounds corresponding to α = 5%) coincide on the instability of the estimated pa-

rameters that describe Colombian commodity trade with China, the Netherlands and the Dominican

Republic. On other cases the combined results of both tests are not conclusive: the CUSUM stabil-

ity test considers unstable Colombian commodity trade with Panama, and the CUSUMSQ considers

unstable that with the U.K. and Switzerland. Colombian bilateral trade of non-commodities with the

considered partners appears to be more stable. Both stability tests did not concurrently considered

unstable the coe�cients of any model describing this type of Colombian trade. Only the CUSUMSQ

test considers unstable such trade with Venezuela. Most model estimations did not give evidence

of serial correlation in the residuals or misspeci�cation. According to the Lagrange Multiplier test

statistic, considering the critical value of χ2 at a signi�cance level of 5% and one degree of freedom,

e.g. χ2α=5% (1) = 3.84, the residuals of the regression estimations for Colombian bilateral commodity

trade with the U.K. and Venezuela and for Colombian bilateral non-commodity trade with Belgium,

China, the U.S., Mexico and the Netherlands show �rst order serial correlation. The Ramsey test

statistic for model functional misspeci�cation, also considering χ2α=5% (1) = 3.84, indicate that the

estimation equations for Colombian commodity trade with Panama, and non-commodity trade with

Italy, Switzerland and Venezuela are misspeci�ed.

6 Conclusion

A dynamic adjustment mechanism is used to analyze the short-run and long-run responses of

Colombian international trade to real exchange rate volatility. This is done under the scope of

the Marshall-Lerner condition, a cointegration analysis and the J-curve hypothesis where an Error-

Correction model is used to estimate Colombian bilateral commodity and non-commodity trade with

its main exchange partners for the 1998-2010 period. The main international trade partners of Colom-

bia in term of total volume of trade are Germany, Belgium, Brazil, China, Ecuador, the U.S., the

Netherlands, Italy, Japan, Mexico, Panama, Peru, the U.K., the Dominican Republic, Switzerland and

15

Venezuela. International trade with these countries represents 80.1% of Colombia's total exports and

72.7% of total imports.

Colombian bilateral trade of commodities with each of its main trade partners is cointegrated.

That is, the commodity bilateral balances of trade for Colombia with all of these countries are in

long-run equilibria with their real exchange rates and their income levels. Cointegration is also always

found in the Colombian bilateral trade of non-commodities with those countries that their national

currency is one of the world's main currencies, e.g. the U.S. dollar, the Euro, the British pound

and the Japanese Yen. Such a precise result is not found on the Colombian bilateral trade of non-

commodities with countries that their national currency is distinct from those that are the world's

main currencies. Considering these results, it may be argued that the distinctions in the presence of

long-run equilibria in international trade may be due to the currency denomination of the trade volume

and not speci�cally the trade partner; i.e. non-commodity trade is likely to be denominated in the

currency of the exporting country, while commodity trade is often priced in terms of the world's main

currencies.

Although no evidence sustaining the J-curve hypothesis was found analyzing Colombian bilateral

trade of commodities or of non-commodities, individual scenarios arouse where either Colombian bi-

lateral net exports initially decrease following a real devaluation of the Colombian currency, or where

Colombian net exports increased in the longer run trailing a devaluation. These are cases where either

the �price e�ect� or the �volume e�ect� following a devaluation is present, but not both. The �price

e�ect� or the short-run decrease in the trade balance after the devaluation is never observed in the

Colombian bilateral trade of commodities, nor in the Colombian bilateral trade of non-commodities

with each of its main Latin-American partners. The �price e�ect� in some occasions is observed in the

non-commodity trade of Colombia with countries that their national currency is one of world's four

main currencies. The opposite e�ect, or a short-run increase in the trade balance following a deval-

uation, is signi�cantly a more predominant behavior of the Colombian bilateral balance of trade of

non-commodities than is a short-run decrease. Colombian bilateral trade, with most of its trade part-

ners and particularly in the trade of non-commodities, experienced long-run increases in its net-export

16

after a devaluation.

Summarizing, the results presented in this paper indicate that the detriment of the Colombian

terms of trade will have no signi�cant short-run impact on the bilateral trade of commodities with

any of its main partners and will have a positive long-run impact on the bilateral trade of these

goods exclusively with the country's main Latin-American trade partners. The satisfaction of the

ML-condition in only a few of these trade scenarios suggests that it is trivial the volume of the balance

of payments of commodity trade that is a�ected by real exchange rate volatility. The detriment of

the Colombian terms of trade will have a negative short-run impact on the bilateral trade of non-

commodities with a considerable number of its main trade partners as the J-curve hypothesis would

predict, but most of the volume of Colombian bilateral trade of non-commodities exhibits short- and

long-run improvements after a real devaluation of the currency. In great contrast with what observed

in the Colombian trade of commodities, the ML-condition suggests that real exchange rate volatility

a�ects a signi�cant volume of the balance of payments from the bilateral trade of non-commodities.

Considering that the most recent economic history of Colombia involves stark real appreciations

of its currency it is important to analyze the trade balance from the perspective of improvements in

the terms of trade of the country and its possible impact on the permanence of the current surplus.

The results of this paper suggest that: (1) improving the terms of trade of Colombia will not signif-

icantly a�ect the short-run surplus of Colombian commodity trade, (2) improving the terms of trade

of Colombia will negatively a�ect, in a trivial amount, the long-run surplus of Colombian commodity

trade, mostly by increasing these import categories from its Latin-American partners, and (3) improv-

ing the terms of trade of Colombia will negatively a�ect the Colombian short- and long-run balance

of trade of non-commodities, increasing the current de�cit. These results support the conjecture that,

ceteris paribus, if continuous improvements in the terms of trade of Colombia are not accompanied by

enhancements in the country's productivity (e.g. international di�erentials in wages and/or unit labor

requirements), especially in the manufacture of non-commodities, then the permanence of the current

Colombian trade balance surplus is not sustainable over a longer horizon.

17

References

[1] Bahmani-Oskooee, M. 1998. �Cointegration Approach to Estimate the Long-Run Trade Elasticities

in LDCs.� International Economic Journal 23 (3): 89-96.

[2] Bahmani-Oskooee, M. & T. Brooks. 1999) "Bilateral J-Curve between U.S. and her trading part-

ners." Review of World Economics (Weltwirtschaftliches Archiv) 135 (1): 156-165.

[3] Bahmani-Oskooee, M. , and G. Goswami. 2003. �A Disaggregated Approach to Test the J-Curve

Phenomenon: Japan versus Her Major Trading Partners.� Journal of Economics and Finance 27

(1): 102-113.

[4] Bahmani-Oskooee, M., and M. Hajile. 2010. �How sensitive is the US-Sweden trade to kronor-

dollar rate: evidence from industry data.� Global Business and Economic Review 12, (4): 359-383.

[5] Bahmani-Oskooee, M. and S.W. Hegerty. 2010. �The J- and S-curves: a survey of the recent

literature.� Journal of Economic Studies 37 (6): 580-596.

[6] Bahmani-Oskooee, M. and S.W. Hegerty. 2011. �The J-curve and NAFTA: evidence from com-

modity trade between the US and Mexico.� Applied Economics 43 (13): 1579-1593.

[7] Beidas-Strom, S., J. Kang, P. Loungani, A. Matsumoto, and M. Rousset �Commodity Mar-

ket Review.� in World Economic Outlook, October 2012. The International Monetary Fund.

http://www.imf.org/external/np/res/commod/index.aspx.

[8] Boughton, J., and W. Branson. 1988. �Commodity Prices as a Leading Indicator of In�ation�,

The National Bureau of Economic Research. http://www.nber.org/papers/w2750

[9] Brown, R. L., J. Durbin, and J. M. Evans. 1975. �Techniques for Testing the Constancy of Re-

gression Relations Over Time.� Journal of the Royal Statistical Society Series B 37 (2): 149-192.

18

[10] Dwyer, J., C. Kent, and A. Pease. 1994. �Exchange Rate Pass-Through: Testing the Small Country

Assumption for Australia.� Economic Record 70 (211): 408-423.

[11] Engle, R. E. and C. W. Granger. 1987. �Cointegration and Error Correction: Representation,

Estimation and Testing.� Econometrica 55 (2): 251-276.

[12] Gomes, F. A. R. and L. S. Paz. 2005. �Can real exchange rate devaluation improve the trade

balance? The 1990-1998 Brazilian case.� Applied Economics Letters 12 (9): 525-528.

[13] Gupta-Kapoor, A. and U. Ramakrishnan. 1999. �Is there a J-Curve? A new estimation for Japan.�

International Economic Review 13 (4): 71-79.

[14] Halicioglu, F. 2008. �The Bilateral J-Curve: Turkey Versus her 13 Trading Partners.� Journal of

Asian Economics 19 (3): 236-243.

[15] Heston, A., Summers, R. and Aten, B. Penn World Table Version 7.1 (Center for International

Comparisons of Production, Income and Prices at the University of Pennsylvania; July 2012).

https://pwt.sas.upenn.edu/php_site/pwt_index.php.

[16] Huertas, C. 2003. �Revisión metodológica del ITCR y cálculo de un índice de competitividad con

terceros países.� Revista del Banco de la Republica 76 (913): 46-73.

[17] Junz, H. B. and Rhomberg, R. R. 1973. �Price Competitiveness in Export Trade among Industrial

Countries.� American Economic Review (Papers and Proceedings) 63 (2): 412-418.

[18] Kremers, J. J., N. R. Ericson and J. J. Dolado. 1992. �The Power of Cointegration Tests.� Oxford

Bulletin of Economics and Statistics 54 (3): 325-348.

[19] Magee, S.P. 1973. �Currency Contracts, Pass-Through, and Devaluation.� Brookings Papers on

Economic Activity 4 (1): 303-325.

[20] Murphy, H. 2010. �Colombia, Venezuela Agree to Restore Diplomatic Relations After Dispute.�

Bloomberg, August 11. http://www.bloomberg.com/news/2010-08-11/colombia-venezuela-agree-

to-restore-diplomatic-relations-after-dispute.html

19

[21] Pesaran, M.H., Y. Shin and R. J. Smith. 2001. �Bounds testing approaches to the analysis of level

relationships.� Journal of Applied Econometrics 16 (3): 289-326.

[22] Roberts, M. and W. Schlenker. 2010. �Indentifying Supply and Demand Elasticities of Agricultural

Commodities: Implications for the U.S. Ethanol Mandate.� The National Bureau of Economic

Research. http://www.nber.org/papers/w15921.

[23] Shahbaz, M., A. Jalil and F. Islam. 2012. �Real Exchange Rate Changes and the Trade Balance:

The Evidence from Pakistan.� The International Trade Journal 26 (2): 139-153.

[24] Sistema Estadistico de Comercio Exterior, SIEX de la DIAN de Colombia

http://websiex.dian.gov.co/pls/siex/balcpais$.startup

[25] �Venezuela and Colombia: Politics versus trade.� The Economist, September 10, 2009.

http://www.economist.com/node/14416724

[26] Wilson, P. 2001. �Exchange Rates and The Trade Balance for Dynamic Asian Economies: Does

the J-Curve exist for Singapore, Malasya and Korea?� Open Economies Review 12 (4): 389-413.

20

Table 4: Appendix A - HS Classi�cation Codes and Description of Colombian CommoditiesHead & Subhead Description of commodity good

0603Cut �owers and �ower buds of a kind suitable for bouquetsor for ornamental purposes, fresh, dried, dyed, bleached,,impregnated or otherwise prepared.

090111 Co�ee, not roasted, Not deca�einated090112 Co�ee, not roasted, deca�einated090121 Co�ee roasted, Not deca�einated090122 Co�ee roasted, deca�einated100510 Cornseed170111 Raw sugar not containing added �avouring or colouring matter. Cane sugar170199 Raw sugar not containing added �avouring or colouring matter. Other2704 Coal

2709Mineral fuels, mineral oils and products of their distillation;bituminous substances; mineral waxes

2710 Petroleum oils and oils obtained from bituminous minerals, other than crude710812 Gold (including gold plated with platinum)720260 Ferro-nickel

21

Table 5: Appendix B - Colombian Non-Commodity Exports and Imports (million USD, 1998-2009)GERM

BEL

BRA

CHIN

ECUA

U.S.

ITA

JAP

MEX

NET

PAN

PER

REIN

U.K.

SWI

VEN

Animal

X1.8

0.8

0.3

0.3

6.0

35.6

1.0

5.4

0.6

0.3

1.4

0.6

0.1

0.2

0.1

287.8

Products

M0.7

0.3

3.1

1.4

13.0

19.6

0.1

0.0

0.6

1.4

2.4

3.5

0.7

0.0

0.2

3.9

Vegetable

X90.6

176.7

6.6

2.3

13.0

203.8

35.6

4.4

14.8

16.5

4.5

6.2

50.9

5.0

1.0

57.5

Products

M3.9

1.2

32.4

10.4

76.5

506.4

1.7

0.3

9.6

9.0

0.0

8.3

1.0

0.0

0.2

18.7

Food

X16.1

6.1

3.7

0.2

64.0

112.7

24.2

16.6

11.9

7.6

8.8

30.3

13.8

15.0

1.4

138.3

Products

M2.7

1.8

58.2

6.5

85.8

136.1

6.7

0.2

29.2

7.5

1.7

33.7

16.7

0.3

1.3

48.8

Mineral

X75.0

16.9

29.7

1.8

54.7

634.9

82.5

6.2

12.7

330.2

16.1

32.8

152.3

28.3

13.7

61.4

Products

M2.6

0.1

9.5

9.2

6.7

18.0

2.0

1.1

2.8

0.3

0.1

2.8

1.6

0.6

1.2

6.3

Chemical

X2.4

3.6

28.1

4.2

186.1

131.2

2.4

1.2

62.8

1.1

54.0

97.8

3.6

19.7

37.8

252.8

Products

M237.6

41.9

132.5

137.0

22.1

1,113.8

64.0

53.7

253.5

45.6

9.1

27.2

65.9

0.2

113.2

208.9

Plastics/

X1.5

1.5

69.0

3.8

104.3

82.3

4.2

0.0

48.5

2.7

13.4

89.3

2.8

27.5

0.3

148.5

Rubber

M43.7

13.5

80.5

68.7

27.6

348.8

19.3

54.0

79.9

5.8

1.8

31.9

12.3

0.0

3.4

95.7

Leathers

X1.7

2.1

0.5

8.9

3.4

39.7

28.3

0.9

8.6

1.0

2.1

1.2

1.0

1.5

0.0

53.6

M0.4

0.1

1.9

24.3

0.7

3.5

1.6

0.1

0.5

0.0

0.5

0.6

0.1

0.0

0.1

0.3

Wood

X0.3

1.1

4.1

1.6

89.9

38.7

0.2

1.7

37.1

0.3

18.5

46.9

3.0

16.7

0.0

158.6

Products

M25.7

3.9

37.3

16.1

25.4

164.2

8.6

1.9

19.9

4.1

0.7

16.4

6.2

0.6

2.4

15.4

Textiles

X13.4

0.2

9.5

0.7

84.5

360.3

5.8

0.2

80.3

0.8

13.9

27.6

8.8

5.9

0.3

388.8

M9.9

4.1

55.3

127.4

32.7

188.2

20.5

2.8

48.8

2.8

4.9

29.8

3.4

0.1

1.9

18.2

Footwear/

X1.9

0.5

0.1

0.0

7.9

5.1

0.0

0.0

2.7

0.3

2.4

1.0

0.2

0.6

0.0

44.1

Headgear

M0.2

0.0

9.3

66.5

16.0

4.4

2.4

0.1

0.7

0.2

10.7

1.8

0.1

0.0

0.0

0.9

Stone

X3.1

2.0

3.9

0.1

31.5

216.3

2.5

15.0

10.4

0.4

9.4

14.4

4.4

5.6

81.5

69.5

&Glass

M6.8

1.2

25.2

36.2

6.5

28.9

8.1

1.6

19.8

0.4

0.8

10.4

1.1

0.0

0.4

11.3

Metals

X1.0

1.5

13.8

70.1

54.9

130.9

1.2

1.5

20.5

11.1

17.2

34.9

0.8

9.9

0.1

108.7

M42.9

5.3

210.5

109.2

31.7

210.6

18.3

99.2

108.4

3.5

3.2

131.1

15.5

0.3

4.0

332.8

Machinery

/X

1.9

0.1

5.2

1.8

66.1

81.2

1.2

0.2

27.5

1.6

14.1

36.8

1.3

8.8

0.3

204.8

Electrical

M290.6

18.1

261.4

614.2

16.1

1,571.3

168.0

219.3

581.3

41.5

17.2

6.0

55.7

0.3

62.7

26.6

Transports

X0.1

0.0

1.7

0.0

119.4

40.2

0.2

0.0

1.9

0.3

4.1

1.8

0.2

1.3

0.0

296.2

M62.9

1.9

150.6

56.3

107.9

786.0

10.9

233.8

192.1

20.3

8.7

0.2

10.2

0.5

0.5

90.7

Miscel.

X1.9

0.4

7.0

0.1

19.4

47.4

0.6

0.6

8.3

0.4

10.2

9.3

1.0

5.2

0.3

61.1

M56.1

1.7

31.5

139.2

1.9

325.7

19.9

45.7

18.5

14.0

5.0

4.1

13.8

0.7

25.8

4.2

Services

X0.1

0.0

0.1

0.0

0.0

0.9

0.0

0.0

0.2

0.0

0.1

0.1

0.1

0.0

0.1

0.1

M0.1

0.0

7.9

25.6

0.1

3.4

7.3

14.3

0.3

0.0

0.1

0.1

0.1

0.0

3.1

0.2

22